If you’ve ever spent a full day chasing a weight reading that won’t settle, you know the exact mix of frustration and paranoia. You wonder if the instrument is lying. You wonder if you’re losing it. Then you find out the problem was a variable-frequency drive two feet away, and you feel stupid.
I’ve been working with precision instruments for about 12 years. I’ve bought them, installed them, calibrated them, and broken them. I’ve also made and documented enough expensive mistakes to know that most measurement problems aren’t the instrument. They’re the gap between instruments. This is a story about that gap, and why knowing your own expertise boundary is more valuable than any brand name.
The surface problem: a “faulty” load cell that wasn’t faulty
In September 2022, a tank scale in our process area started drifting. The display would settle at 12,000 kg, then creep to 12,020 kg, then fall back. We cleaned the junction box, checked the grounding, and swapped the transmitter. No change. The conclusion, after two days, was that the METTLER TOLEDO load cells on that tank were bad.
I went back and forth for two days between replacing the load cells and calling an outside contractor. Replacing them was easy to justify. Calling a contractor meant admitting that my team had hit a wall. The numbers said the load cell signal was in the normal range. My gut said something else was wrong. I finally called the contractor.
He showed up with a Tektronix oscilloscope. Within 20 minutes, he found a burst of noise that happened exactly when the mixer’s variable-frequency drive (VFD) ramped up. The shield on the signal cable was damaged at a junction box, so the VFD noise was riding on the millivolt output. The load cells were fine. The real cost wasn’t the $3,200 replacement quote. It was the seven days of downtime, the $900 contractor invoice, and the embarrassment of having written “load cell failure” in the work order.
The deeper cause: a blind spot between instrument families
Most of us are trained inside a single instrument community. Lab people know balances, pH meters, and spectrophotometers. Process people know load cells, weighing terminals, and transmitters. Electricians know power distribution, multimeters, and oscilloscopes. But a measurement chain often crosses all those communities, and nobody owns the whole chain.
Take load cells. A METTLER TOLEDO load cell’s output is a small DC millivolt signal—typically 2 or 3 mV per volt of excitation. (Don’t hold me to the exact number; it varies by model, so check the datasheet.) That means a full-scale output is often 20 to 30 millivolts. If you check that with an electrician’s digital multimeter, you might see “about 25 mV” and think the signal is fine. But that reading could be misleading, because the meter’s input circuit wasn’t designed for precise low-level DC measurements near a machine with variable-frequency drives.
To be fair, an electrician’s digital multimeter is a reliable workhorse for its real job: troubleshooting 24 VDC loops, checking 480 VAC phases, verifying continuity. I would never tell an electrician to stop carrying one. But when you’re looking for microvolt-level drift on a load cell, you need a meter with suitable resolution and accuracy on the low millivolt range. Most multimeters are good at some things and not others; that’s not an insult, it’s a design trade-off.
I’m not an electrical engineer, so I can’t speak to the input impedance details of every multimeter model. What I can tell you from a specifications standpoint is this: check the accuracy on the actual mV range before you trust a reading. If the meter’s specifications say 1.5% of reading plus a few digits on that range, a 20 millivolt signal can easily be off by 0.3 millivolts or more. For a load cell that only puts out 30 millivolts at full scale, that’s roughly 1% of the output—enough to hide real drift.
This gets into electrical metrology territory, which isn’t my expertise. I’d recommend consulting an instrumentation engineer before you buy a meter just for load cell testing. But I can tell you from experience: don’t assume your electrician’s digital multimeter has the precision you need. Check the numbers.
The same boundary problem shows up in the lab
We also use a METTLER TOLEDO UV-Vis spectrophotometer in our lab. It’s a beautiful instrument when the sample is prepared correctly. But “correctly” is the key part. In March 2023, a chemist complained that the UV-Vis results were nonsense. The absorbance values were too high, and the spectrum had weird spikes. He blamed the spectrophotometer.
It wasn’t the spectrophotometer. The sample had undissolved particles and a slight haze. A UV-Vis spectrophotometer measures the light transmitted through a clear solution. If you give it a cloudy suspension, scattered light makes everything look like it absorbs too much. The instrument was doing exactly what it was designed to do. We were asking it to be a turbidity meter because we didn’t feel like re-filtering the sample.
Again, the instrument wasn’t broken. Our process was. If we had called the vendor and insisted the UV-Vis spectrophotometer was broken, we’d have been wrong—and we’d have looked foolish after they demonstrated the problem in five minutes.
The cost of pretending we know it all
The obvious cost is the replacement part you don’t need. I nearly spent $3,200 on a load cell that was healthy. I once watched another team replace a pH probe and transmitter before discovering a bad connector. Those things add up.
But the bigger cost is credibility. When your team writes “instrument failure” on a work order and it isn’t a failure, you train the next person to distrust your judgment. The next time you say “this is serious,” they may not listen. That’s hard to put a price on, especially in a plant where trust is the real currency.
In a regulated environment, the cost shows up in audits. According to ISO 9001:2015, Clause 7.1.5, you must provide the right equipment for monitoring and measurement—and that includes test equipment used to verify production instruments. If you can’t show that your electrician’s digital multimeter has enough resolution to verify a load cell signal, an auditor can question the entire calibration record. To me, that’s a far bigger risk than any single $3,200 part.
The only solution I’ve found: respect the boundary
I’m not against cross-training. I’m against the idea that one person or one brand can be the world’s authority on everything. I’d rather work with a specialist who knows their limits than a generalist who overpromises. The vendor who says “this isn’t our strength—here’s who does it better” sounds more honest than the one who says “we do it all.”
After the VFD noise incident, I made a simple checklist for our team. It’s not a complete system, but it catches most of the embarrassing mistakes:
- Name the signal first. Is it a millivolt load cell signal, a 4-20 mA process loop, a 0-10 V output, or a fiber optic signal? Don’t pick a test tool until you know.
- Match the test tool to the signal. If you’re verifying a precision load cell, use a meter with documented accuracy on the low mV range, or call someone who has one. An electrician’s digital multimeter is for electrical work—not necessarily for metrology.
- Ask who owns the boundary. If you suspect noise from a VFD motor, that’s an electrical engineering problem. Get someone who knows how to use a Tektronix oscilloscope, or whatever tool is appropriate, and who understands grounding and shielding.
- Write the observation, not the conclusion. Instead of “load cell faulty,” write “reading unstable; signal shows periodic noise bursts; cause unknown.” That small change saves other people from following your assumption.
I also added one rule to our purchase process: before ordering a major replacement part, we have to document the observed signal and explain why our test method was trustworthy. It takes ten minutes. It has already stopped two unnecessary orders this year.
If you’re dealing with a Mettler-Toledo load cell, a METTLER TOLEDO UV-Vis spectrophotometer, or any precision instrument, the instrument is usually more honest than you think. The human decisions around it are where the uncertainty comes in.
Trust me on this one: the most valuable skill you can build isn’t troubleshooting every instrument in the plant. It’s knowing when to call somebody else.
Leave a comment